Strongly correlated systems

in atomic and condensed matter physics

Lecture notes for Physics 284

by Eugene Demler
Harvard University

September 18, 2014

2
Chapter 5

Atoms in optical lattices

Optical lattices provide a powerful tool for creating strongly correlated many-
body systems of ultracold atoms. By choosing different lattice geometries one
can obtain very different single particle dispersions. The ratio of the interaction
and kinetic energies can be controlled by tuning the depth of the lattice.

5.1 Noninteracting particles in optical lattices

The simplest possible periodic optical potential is formed by overlapping two
counter-propagating beams. Electric field in the resulting standing wave is

E(z) = E0 sin(kz + θ) cos ωt (5.1)

Here k = 2π/λ is the wavevector of the laser light. Following the general recipe
for AC Stark effects, we calculate electric dipolar moments induced by this field
in the atoms, calculate interaction between dipolar moments and the electric
field, and average over fast optical oscillations (see chapter ??). The result is
the potential

V (z) = −V0 sin2 (kz + θ) (5.2)

where V0 = α(ω)E02 /2, with α(ω) being polarizability. It is common to express

V0 in units of the recoil energy Er = ~2 k 2 /2m. In real experiments one also
needs to take into account the transverse profile of the beam. Hence V (r⊥ , z) =
exp{−2r2 /w2 (z)} × V (z). In most experiments the main effect of the transverse
profile is only to renormalize the parabolic confining potentia. Combining three
perpendicular sets of standing waves we get a simple cubic lattice

V (r) = −Vx 0 cosqx x − Vy 0 cosqy y − Vz 0 cosqz z (5.3)

3
4 CHAPTER 5. ATOMS IN OPTICAL LATTICES

Figure 5.1: Different geometries of optical lattices. Two perpendicular standing

waves make a two dimensional lattice, which confined atoms into tubes. Three
standing waves make a three dimensional lattice. Figure taken from ??.

5.1.1 Band structure

Let us now discuss solutions of the single particle Schroedinger equation in such
periodic potential
~2
H=− (∇2 + ∇2y + ∇2z ) + V (r) (5.4)
2m x
For potential (5.3) we can separate variable in the single particle Schroedinger
equation
Ψ(x, y, z) = Ψ1 (x)Ψ2 (y)Ψ3 (z) (5.5)
For each of the coordinates we have a Matthieu equation
~2 d2
(− − V0 cos qx x) Ψ1 (x) = E Ψ1 (x) (5.6)
2m dx2
Eigenevalues and eigenfunctions of the Matthieu equation are known. We will
mostly be interested in the regime of a deep lattice, known as the tight-biding
regime.
In the tight binding regime, dispersion in the lowest Bloch band is given by
qx = −2J cos qx d
 3/4  
4 V0 V0
J = √ Er exp −2( )1/2 (5.7)
π Er Er
Here d = 2π/q is the period of the lattice, which is half of the wavelength of
laser light (remember that q = 2k). In subsequent notes we will often set d = 1
5.1. NONINTERACTING PARTICLES IN OPTICAL LATTICES 5

Figure 5.2: Dispersion in a one dimensional potential V (x) = V cos qx. Energy
is units of recoil energy Er = ~2 q 2 /8m. Figure taken from ??.

Figure 5.3: Schematic representation of the tight binding regime. Atoms are
restricted to occupy the lowest Bloch band only. They can hop between nearest
neighbor sites.

to shorten the notations. The Hamiltonian for free motion in the lowest Bloch
band
X † X
H = −J bi bj = q b†q bq (5.8)
hiji q

Here hiji are nearest neighbors.

5.1.2 Experimental tests of single particle dispersion in

optical lattices
Figures (5.4), (5.5),(5.6) show some of the experiments used to probe band
structure of non-interacting particles in an optical lattice. Note a remarkable
possibility to change the band structure in time, e.g. to turn the lattice on and
off.

5.1.3 Semiclassical dynamics

Semiclassical dynamics of noninteracting particles in a periodic potential nd in
the presence of external fields can be reduced to three simple rules.

• Band index is constant. This rule works as long as forces are not too
strong. For strong forces interband tunneling becomes possible.
6 CHAPTER 5. ATOMS IN OPTICAL LATTICES

Figure 5.4: Quench experiments with bosonic atoms in an optical lattice demon-
strate decomposition of band eigenstates into plane waves. Optical lattice was
switched on suddenly. Initial state is a state with a well defined physical mo-
mentum p = 0. When the optical lattice is on, this is no longer an eigenstate of
the Hamitlonian, but a superposition of states from different bands with quasi-
momentum p = 0. Different components of the wavefunction start to evolve ac-
cording to their energies. After some evoltuion the optical potential is switched
off abruptly again. The final state is a superposition of states with different
physical momenta that differ by reciprocal lattice vectors. Weights of different
momentum components in the final state are measured using TOF technique.
Figs. taken from [1].

• d~
r
dt = ∂
∂~
k
k
= ~vg . This is the usual rule of the wave dynamics. Propagation
of a wavepacket is controlled by the group velocity.
d~
• k
dt = F~ . This is basically Newton’s second law.

One of the most striking manifestations of the semiclassical dynamics of

atoms in an optical lattice is Bloch oscillaitons. In the presence of constant
external field, atoms perform oscillations but do not move on the average. We
discus one dimensional system for simplicity. When the force is constant

k(t) = k(0) + F t (5.9)

Using the tight binding limit with k = −2J cos kd and vg = 2Jd sin kd, we
obtain

ẋ = 2Jd sin [(k0 + F t)d]

2Jd
x = x0 − cos [(k0 + F t)d] (5.10)
F
Bloch oscillations have been predicted theoretically for electrons back in the
thirties. Numerous attempts to observe them in solid state systems failed due
to unavoidable disorder in such systems. In the first observation of of Bloch os-
cillations by Dahan et al. [2] only several cycles could be observed. Interaction
between atoms was causing decoherence. In the latest experiments[3, 4] inter-
action contact was switched off using Feshbach resonance and tens of thousands
of Bloch oscillations could be observed.
5.1. NONINTERACTING PARTICLES IN OPTICAL LATTICES 7

Figure 5.5: Lattice modulation experiments with bosons were used to probe
single particle dispersion. First lattice acceleration was used to give desired
quasimomentum to atoms. Then modulation of the lattice was used to excite
atoms into the higher band. Atoms were excited when the frequency matched
the energy difference between the bands. Note that lattice modulation can only
excites atoms into states of the same symmetry in individual wells. Thus we
only se transitons n = 0 → 2, 4, ... Figs. taken from [1].

Figure 5.6: Noninteracting fermions in an optical lattice. TOF experiments

measure occupations in momentum space. With increasing density the shape of
the Fermi surface changes. At the lowest density we see spherical Fermi surface.
At the largest density the entire first Brillouin zone is filled (square). Figures
taken from [5].

5.1.4 State Dependent Lattices

It is possible to use selection rules for optical transitions to make different lattice
potentials for different internal states of atoms. As an example, we consider
an atom with the fine structure as for 23 Na and 87 Rb, interacting with two
circularly polarized laser beams (see fig5.8). The right circularly polarized light
σ + couples |S = 1/2, ms = −1/2i to two excited levels P1/2 and P3/2 and
detunings of opposite signs (see figure 5.8). AC Stark effects due to P1/2 and
P3/2 have the opposite signs and cancel each other for appropriae frequency ωL .
Thus at ωL the AC Stark effect of the |S = 1/2, ms = −1/2i will come only from
σ− polarized light, which we denote as V− . Analogously |S = 1/2, ms = +1/2i
will only be affected by the σ+ , which gives us the potential V+ (x).

Decomposing states | F = 2, mF = 2 i and | F = 1, mF = ±1 i into states

8 CHAPTER 5. ATOMS IN OPTICAL LATTICES

Figure 5.7: Bloch oscillations in cold atoms. The left figure shows physical
momentum as a function of time. At first, atoms get accelerated from k =
0, but when they approach the edge of the Brillouin zone, they undergo a
Bragg reflection. Note that the Bragg reflection appears only very closely to
the Brillouin zone edge. This is a property of the weak potential. The right
figure shows velocity overaged over the original and the Bragg reflected parts of
the cloud. Figures taken from [2].

with certain electron spin ms , we find

V| F =2, mF =2 i = V+ (x)
3 1
V| F =1, mF =+1 i = V+ (x) + V− (x)
4 4
1 3
V| F =2, mF =2 i = V+ (x) + V− (x) (5.11)
4 4
5.1. NONINTERACTING PARTICLES IN OPTICAL LATTICES 9

Figure 5.8: Use of light polarization to make spin dependent optical lattices.
Atom is assumed to have one electron in the outer shell. In the ground state
the outer shell electron of the atom has two hyperfine states S1/2 , i.e. angular
orbital momentum l = 0, electron spin ms = ±1/2. In the excited state electron
has orbital angular momentum l = 1. Fine structure splitting separates the
manifold of the excited states into states with total angular momenta 1/2 and
3/2. AC Stark effects due to P1/2 and P3/2 levels have the opposite signs and
cancel each other for the appropriate frequency ωL .
10 CHAPTER 5. ATOMS IN OPTICAL LATTICES

Figure 5.9: These experiments by O. Mandel et al. demonstrate making spin de-
pendent optical lattices using light polarization and creation of entangled atomic
states over many lattice sites. One dimensional optical lattice is formed from
two counter-propagating laser beams with linear polarizations. Polarization of
the returning beam is controlled by the electro-optical modulator. When polar-
ization angles enclose vector θ, effective potentials for atoms with ms = ±1/2
are V+ (x) = Vmax cos2 (kx+θ/2) and V− (x) = Vmax cos2 (kx−θ/2). Correspond-
ing potentials for hyperfine states of the atoms are given in equation (5.11). The
second figure shows interferometer sequence used to delocalize an atom over ar-
bitrary number of lattice sites. In the third figures interference patterns of the
TOF experiments demonstrate coherence over many sites. Period of the inter-
ference pattern is λinterference = ht/mnd, where n is the number of sites over
which an atom was delocalized, d is the OL period, t is the expansion time, m
is the mass of atoms. Figures taken from [6].
Bibliography

[1] J. Hecker Denschlag et al. J. Phys. B: At. Mol. OPt. Phys., 35:3095, 2002.
[2] M. Dahanl et. al. PRL, 76:4508, 1996.

[3] M. Dahanl et. al. PRL, 100:80405, 2008.

[4] M. Gustavssonl et. al. PRL, 100:80404, 2008.
[5] M. Kohl et. al. PRL, 94:80403, 2005.

[6] O. Mandel et. al. PRL, 91:10407, 2003.

11